Skip to content

Tracking: .svar includes variants past the region end that .svar2 clips (6/1000 rows on 1kg test data) #429

Open

Description

@d-laub

Summary (tracking, not urgent)

For the same regions and samples, .svar and .svar2 datasets can return different variant sets. .svar includes a few variants 6–55 bp past the region end; .svar2 clips at the region end.

Seen while adding .svar2 flat output (#428), on the repo's 1kg test data (100 regions of 10 kb, 5 samples): 6 of 1000 (region, sample, ploid) rows differ. Ragged "variants" output shows the same difference, so it predates #428.

On the 994 rows where the variant sets agree, all fields, flank_tokens included, are byte-identical across backends.

Open questions

How to see it

Compare ds.with_seqs("variants") output from .svar and .svar2 datasets written from the same normalized VCF over the 1kg test regions, and diff the per-row variant starts. The parity tests added in #428 compare only the rows where the sets agree.

No fix planned yet; this issue tracks the difference.

🤖 Generated with Claude Code

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions